Improving Ensemble Filters with Flow Matching
Abstract
Data assimilation estimates a dynamical state from partial and noisy observations. Classical ensemble filters are efficient but restrict analysis updates through finite sample covariance and affine Gaussian distribution. We introduce the Flow Ensemble Filter (FlowEF), which uses conditional flow matching to transport the forecast ensemble from a classical baseline filter to an analysis ensemble. FlowEF uses a localized Gaussian source during training, transports forecast ensemble members from a baseline filter at deployment, and conditions its velocity field on ensembles from that baseline filter and the observation. The proposed model therefore learns a nonlinear update while mapping each baseline ensemble independently. For sparsely observed dynamical systems, FlowEF improves both deterministic and probabilistic metrics over all four classical ensemble filters. It also achieves the best performance among the state-of-the-art generative data assimilation models.
Related
- Time-uniform accuracy of ensemble Kalman filters with localization
- On the Oja-Flow-Based Low-Rank Approximation of Kalman-Bucy Filters for Linear Time-Varying Systems
- Multi-Fidelity Particle Flow Filters based on Log Homotopy
- Weighted Adjoint Matching: Finetuning Flow Matching Models with Time-Dependent Control Cost
- Generalized Wasserstein Flow Matching: Nested Couplings for Point Cloud Generation
BibTeX
@misc{chen2026improving,
title = {{Improving Ensemble Filters with Flow Matching}},
author = {Haoyuan Chen and Alexandre Thiéry},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2609.28015},
eprint = {2609.28015},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2609.28015},
}
RIS
TY - UNPB
TI - Improving Ensemble Filters with Flow Matching
AU - Haoyuan Chen
AU - Alexandre Thiéry
PY - 2026
DA - 2026/09/23
PB - arXiv
DO - 10.48550/arxiv.2609.28015
UR - https://arxiv.org/abs/2609.28015
AB - Data assimilation estimates a dynamical state from partial and noisy observations. Classical ensemble filters are efficient but restrict analysis updates through finite sample covariance and affine Gaussian distribution. We introduce the Flow Ensemble Filter (FlowEF), which uses conditional flow matching to transport the forecast ensemble from a classical baseline filter to an analysis ensemble. FlowEF uses a localized Gaussian source during training, transports forecast ensemble members from a baseline filter at deployment, and conditions its velocity field on ensembles from that baseline filter and the observation. The proposed model therefore learns a nonlinear update while mapping each baseline ensemble independently. For sparsely observed dynamical systems, FlowEF improves both deterministic and probabilistic metrics over all four classical ensemble filters. It also achieves the best performance among the state-of-the-art generative data assimilation models.
ER -